Buzz surrounding AI in construction has remained strong over the last few years. From the different tools teams should be using to how to handle data securely, there's no shortage of opinions and options. GCs, in particular, areunder pressure to boost productivity and reduce risk, all while grappling with labor shortages.

While there's plenty of buzz out there, oftentimes, the best way to gain insights into AI best practices is to learn from those who are in the thick of it.

In a recent webinar, AI Tools for General Contractors, I was joined by Brian Athey, Director of Construction Innovation at Miron Construction to learn more about the company’s strategic approach to AI adoption so far.

From practical examples and candid insights, our discussion centered on how AI can solve everyday construction problems. Let's explore some of the top takeaways from our conversation.

Before we get into the specifics of Miron’s AI journey, let’s take a look at some of the top ways AI is being applied today in construction. AI is stepping up to solve some of construction's most pressing pain points. Consider the following.

Inefficiency reduction - Today, construction pros spend too much time trying to find the right data or collect project information. In fact, according to the recent Autodesk 2025 State of Design & Make Report: Spotlight on Construction, construction professionals spend an average of 13 hours per week looking for the right data. AI can help address this by quickly surfacing answers buried deep in drawings, specs, or submittals.

Data fragmentation - Even when data exists, it's often not clean and it's not accessible. This leads to miscommunication and site errors, which ultimately results in rework. AI helps bridge the gap by connecting siloed information, standardizing how teams' access and interpret data, and keeping everyone aligned with real-time updates.

Risk management - Keeping projects on schedule and within budget is table stakes—but it's not always easy. That's why more teams are leaning on AI to proactively flag issues, identify bottlenecks, and ensure smarter, faster decision-making before costly problems arise.

AI holds a lot of promise for GCs, but how can firms actually implement artificial intelligence in their organizations? And what results have other GCs seen once they've taken the plunge?

To that, Brian Athey shared his firsthand experiences and lessons from testing, implementing, and iterating with AI.

According to Brian, Miron didn't start with a dedicated AI department or flashy pilot. They began with the people who already understood the company's technology stack.

"The very first thing that we did was put a committee together. We basically turned our innovation committee into our AI committee. So, the same people who deal with technology across the company were tapped to make decisions on AI tools."

Exploring the concept of a data lake

From there, Miron tackled one of the most significant barriers to AI adoption: fragmented data.

"I'm thankful that we started when we did, and that was when we started looking at the concept of a data lake. Just like many other GCs our size, we've got data scattered in many different areas, and it was important to start bringing all that data into one data lake."

For Miron, this means unifying data from various systems, including the firm's ERP, project management software, CRM, HR, and more. The goal, said Brian, was to "create a clean, secure, permissioned data set to work from."

And while the data lake is still a work in progress, the team is making real strides.

"We actually hired two data analysts to do work on that. We probably could use 10 more, but we got rolling"

Training is another big piece of the AI puzzle. Miron's training initiatives are still evolving, but the focus remains clear: to provide end users with the clarity and confidence to use AI responsibly.

"It's all about figuring out the actual training and message to the end user. What they can do, what they can't do, how they should search, as well as what they can and can't put out there."

A lot of these components are addressed in Miron's internal AI policy, which is intentionally broad to stay adaptable over time.

"The policy doesn't have to be very specific. In fact, we didn't get very specific on purpose because we know technology is going to change."

The policy emphasizes a human-centric approach—making sure AI enhances, not replaces, people's capabilities. It also aligns with existing HR principles, which outline that AI must be used lawfully, ethically, and transparently, and only when necessary.

As for implementation and results, Brian emphasizes that AI delivers value only when it solves real problems. At Miron, the team prioritizes tools that drive efficiency and reduce manual effort. Here are some of the solutions Miron has implemented and the results that teams have seen.

One area where Miron has seen massive time savingsis in submittal log creation. Thanks to AutoSpecs what used to take a week now takes a matter of hours.

"We immediately found 75% savings in creating our submittal logs. In the past, our document controls administrator (DCA) would go through the specification manual and literally create an Excel spreadsheet of all the submittals and their requirements and then send that off to a project manager for approval before creating a submittal log."

According to Brian, this manual process represented about a week's worth of work per project. Now, AutoSpecs analyzes specification documents, allowing users to generate accurate submittal registers and understand project requirements quickly.

"You have your initial results in literally two to five minutes, and then they're reviewing it for maybe the rest of the day going through and kind of cleaning it up and sending and creating the log. So, you're talking four days saved out of five."

Another area where AI is proving its worth at Miron is during preconstruction—specifically, in the estimating process.

The team has been using Autodesk Assistant, powered by Autodesk AI, a generative AI tool that enables users to query large, complex documents, such as specifications documents, to surface specific requirements.

"Most of the value initially would be in our estimating department. They're able to upload the specifications during the estimating process, and they can query concrete requirements, quality requirements, commissioning requirements, compression, etc. All of that has saved a tremendous amount of time."

To formalize the process and ensure consistency, Miron's estimating team is developing a standardized list of AI prompts they plan to ask on every project.

"Our estimating department is currently in the process of creating the 25 must-ask questions on every project… There's so much value here that we needy to create a list and make it best practice."

Regardless of what tools you're using and how you're using them, it's crucial to approach AI with flexibility and a long-term mindset. Here are a few of Brian recommendations.

Recognize that the main platforms you're using will continue to embed AI functionality. Brian recommends being strategic with third-party tools, knowing that AI capabilities are rapidly being integrated into existing software stacks.

"One thing to think about when testing your tools is that what you see today is probably going to be embedded in most of your main tools tomorrow," said Brian.

For example, Autodesk continues to embed AI across its platform; other major systems, like ERPs, are following suit. That's why Brian encourages GCs to test new tools thoughtfully but avoid overcommitting too early.

"I caution making long-term commitments at this point. Keep your commitments short, pilot one to two projects, and go from there."

Strive to have a central data platform. When implementing AI in construction, a centralized, connected data environment is a must. Without it, teams often rely on disconnected point solutions that serve only one project phase or department—and that can limit their impact across the organization.

That's why Autodesk encourages contractors to take a platform-wide approach to AI, where tools are trained on your actual project data and workflows.

What does success look like with AI? That depends on the use case, team, and your overall objectives. Sometimes, you can measure things through metrics like cost savings or ROI. In other cases, the best early indicator that something's working is how quickly it spreads across your teams.

Team enthusiasm can speak volumes. At Miron Construction, Brian said they haven't yet established a complete set of performance metrics, but that hasn't stopped them from gauging impact.

They're paying close attention to usage patterns and organic demand. If a tool solves a real problem, word gets around. "In construction, if a tool is working well, it's like wildfire. It catches on. Teams want it. If it's not, it dies off."

"Teams want it turned on because they heard the other team is doing it, and they got excited because they saw somebody else doing it. To me, that's how I'm measuring success today."

Beyond user adoption, time savings, and data insights, AI is delivering something that's often underestimated: a better day at work. For Brian, the most rewarding part of AI adoption isn't just efficiency—it's seeing people light up when they realize how much easier their job has just become.

"The thing that I think drives us is when you see that light bulb go off and you see somebody smile because they just realize that what you showed them is going to make their life so much simpler."

What's next for AI in construction? Brian admits the pace of change makes it tough to predict more than a few years out. Still, there are three areas he's watching closely.

Brian is excited about multimodal AI—the ability to process and compare different types of data simultaneously, like text and drawings.

"Combining the large language model search of a specification with a drawing and seeing if they sync up, finding errors between the two… I see that coming real soon."

Another trend worth exploring is the integration of AI with automation. Instead of simply surfacing information, tools will begin to take action on your behalf, turning queries into tangible outputs.

"What starts as a simple query could turn into 'now create an RFI for me,' or 'send that email to such and such.'"

Brian sees this shift happening soon, and he's particularly excited about the role of digital agents that not only understand context but can execute tasks.

Finally, Brian is looking forward to how Autodesk Construction Cloud (ACC) evolves into a more unified, predictive platform. Today, ACC helps teams manage essential workflows—like specs, RFIs, and issues—but the future lies in a connected system that can anticipate needs and guide users step-by-step.

"I see the whole Build platform coming together as one… being able to surface data and then act on it."

"I think about somebody who draws every day… I could see a day where the tool surfaces your next click for you, thus making us faster, quicker, more efficient."

Miron Construction's journey shows that implementing AI in construction isn't about chasing trends. It's about solving real problems, empowering teams, and preparing for what's next. From centralizing data to testing the right tools, their experience offers a practical blueprint for GCs ready to take the leap.

Another “buy my AI construction product for marginal impact”. I found it funny this company changed its innovation team to the AI team because organizing data and tracking submittals has been a huge boom on company profits. Give me a break!! This type of AI application is marginal at best. The submittals still need to be reviewed and approved by the project stakeholders. Plus the time you gain with this software can be lost just as quickly with a bad weather event or a backlog on materials. The AI software gods are trying to hard to show how AI is having an impact on overall construction workflows but I just don’t see it. LLMs still have issues and these companies never do a cost benefit analysis on these AI implementations!